摘要
图像语义检索是当前图像检索的一个新的研究热点,其中情感语义是最高层的语义,通常用形容词来描述。本文提出了一种新的图像情感语义注释算法,以图像的简单特征来描述图像内容,建立图像的特征空间;用形容词表达图像的情感语义,通过认知心理实验和因子分析建立情感空间;采用支持向量机算法建立图像内容和其所表达的情感语义之间的联系,并自动对未曾评价过的图像进行注释。在此基础上,提出图像的情感检索方法,实现了服装和风景图像的情感检索,取得了较好的实验结果。
Semantic image retrieval has been a hot research in recent years. Emotion semantic, the highest semantic of an image, is often expressed by adjectives. In this paper, a new emotion semantic annotation algorithm has been proposed. Simple features are used to describe the content of the image and thus construct the feature speace, while cognitive psychological experiments and factor analysis are applied to construct an orthogonal emotional space. Then support vector machine for regression is used to build the mapping function from the image content to its emotion semantic, and automatically annotate unevaluated images. After that, an emotion semantic image retrieval algorithm is presented. Furthermore, emotion scenery and costume image retrieval systems are obtained. At last an interesting experimental result is presented.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第1期27-33,共7页
Pattern Recognition and Artificial Intelligence
基金
国家973计划(No.G1998030500)
中国科学技术大学校青年基金资助项目
